Dissertation Title:  "Policy and Methods in Health Services Research"

This dissertation consists of two policy papers and one methods paper, all grounded in applied, empirical health-services research. The first two papers concern the influence of medical underwriting in the market for long-term care insurance in the U.S. The third presents a framework for evaluating provider preference as an instrumental variable in comparative effectiveness research with multiple treatments, using as an empirical demonstration a study comparing the safety of five atypical antipsychotics and their effects on chronic disease incidence.

In the first paper, I estimate an empirical model of the factors on which firms make decisions to underwrite individuals for long-term care insurance, using data on the health and coverage decisions for applicants at two U.S. firms. I apply the model parameters to a population-based sample to determine what proportion of households of prime ages to purchase long-term care insurance would be able to qualify for policies if they were to apply for coverage. Among the general population, I estimate that 40 percent of individuals would have their applications rejected if they were to apply for long-term care insurance--a rejection rate substantially higher than the rejection rate of 20-25 percent of applicants in the actual market.

The second paper examines policy ramifications of the bounds on coverage in the individual market for long-term care insurance that I establish in paper one. I study two types of policies designed to encourage individuals to purchase long-term care insurance, tax incentives and state Partnership programs, estimating how the effects of differ with respect to individuals’ underwriting probabilities. I exploit variation in the timing of states’ implementation of these policies with difference-in-difference models, and estimate the demand elasticity of long-term care premiums using simulated statewide marginal tax prices as in instrumental variable for individual prices. I find that the response to these policy incentives is highly dependent on individuals' underwriting probabilities: conditional on wealth and income, tax and Partnership have no apparent effect on insurance purchase among low-approval households, and program effect appears to be concentrated among healthier individuals with high approval probabilities. In evaluating reforms for long-term care financing and their potential to increase private insurance rates, as well as reduce financial pressure on public safety-net programs, policy makers need to consider the role of underwriting in the market for long-term care insurance.

The third paper proposes a framework for assessing provider preference as instrumental variables in comparative effectiveness research and describes diagnostic tools to validate (or debunk) a candidate instrument. In an applied analysis, I show how to use these tools to compare multiple treatments. I compare the safety of commonly prescribed atypical antipsychotics using physician prescribing preference as an instrumental variable. Widespread adoption of a basic protocol and road map for validating potential instruments, particularly the use of sensitivity tests and compliers analysis, would improve the quality of comparative effectiveness research.